| With the rapid development of Internet in the world, Internet media is recognized as the fourth media, after newspaper, radio, television. Internet has become one of the significant carriers of network consensus. Social network is an online platform for sharing information, where its novel network structure and interactions bring complexity and fragmentation to online opinions. And it is really a force to be reckoned with network consensus. Social network provides convenient access and flexible interactions, as well as bring out the autonomy of users, complexity of interactions and heterogeneity of the population. The formation, evolution and dissemination of network consensus rely on individual interactions. However, traditional models and research methods of public opinion hardly reflect the realistic characters of social network users accurately. As a result, a series of challenges on network consensus are highlighted increasingly, and point to individual interactions and opinion evolution. The individual interactions require to be further studied from multi-perspectives and small-granularity, such as relationship network structure of users and selection preferences of interactions. The influence of users and their influential ranges should be considered in the process of opinion evolution. Besides, the researches of information dissemination need to highlight the diversity and competition among opinions and information published with the free will of individuals.Consequently, integrating the interdisciplinary research ideas and methods and focusing on the individual interactions of social network, we present studies on data collection and empirical analysis of user behavior, modelling individual interactions, opinion dynamics and evolution, and the pattern of competitive information dissemination in social network. Combined with empirical data analysis, mathematical modeling and computer simulation, it is aimed to discover microscopic rules of individual interactions, to interpret macroscopic phenomenon of opinion formation and evolution, and to explore the dynamics of network consensus. The research is not only able to deeply understand the nature of complex systems and networks consensus, to complete the theoretical studies of the interaction behaviors of individuals, network opinion dynamics and information and consensus dissemination in social network, but also helpful for timely responding and guiding network consensus. The work of the dissertation is supported by the National Natural Science Foundation of China under Grant 61271308 and 61401015, the Beijing Key Laboratory of Communication and Information Systems and the Key Discipline Project of Beijing Education Commission. Main contributions and innovations of the dissertation are as follows:1. Unidirectional following relationship is a significant difference of network structures, between microblog and other social networks, which changes the way individuals getting and spreading information. Empirical analyses are proposed to discover the impact of unidirectional relationships on the network structures, individual interactions, and user preferences of microblog. An integrated data crawling framework on microblog is utilized to collect a large number of user data. It is found that unidirectional relationships are the main reasons causing the distribution of node degrees deviated from power-law. The proportion of bidirectional relationships is much lower than that of unidirectional relationships and decays exponentially. And obvious interaction preferences are observed. Reposting behaviors of both unidirectional and bidirectional friends present power-law characteristics with user preferences. Although its low density, bidirectional relationships play important roles on individual interactions, which have high frequency of interactions and high priority of preferences. The research is helpful for better understanding characteristics of user relationships and behaviors in social network, and provides theory basis and data support for studies on individual interaction of network consensus.2. Individual interactions have great impacts on the formation and propagation of network consensus. A number of studies have shown the power-law and multimodal characteristics in human complex actions; however, mathematical models and descriptions are seldom proposed. In order to finely reproduce and study the complex individual interactions of social network and their time characteristics, an individual interaction model is built. Based on empirical analyses of selection preferences of microblog users, we suggest a preference prioritized decision making mechanism of interested information, and propose the user interaction model based on selection preferences in social network. Then, the influences of observation range and interested probability on individual interactions are analyzed, and the fitness of proposed model is validated by microblog data. It is found that different interaction environments lead to different interaction patterns. Fixed observation brings out exponential distribution of individual interactions, while random observation approximately corresponds to power-law characteristics. We derive mathematical equations of individual interactions to describe the multimodal characteristic and quantify the direct impact of user preferences when facing massive information stream. Further studies on randomization degrees of observed interested information interpret the formation of stationary, decay, and truncation phases. The research is able to reproduce realistic interactions in social network, and give a clear insight into microscopic interaction laws.3. Traditional opinion dynamics models usually make use of simple interaction rules between adjacent individuals, but they are hard to precisely describe opinion formation and evolution in social network. Therefore, we model opinion dynamics for social network in order to investigate the effects of expanded observation and individual heterogeneity on opinion evolution. A novel opinion dynamics model based on expanded observation ranges and individuals’ social influences is proposed. Combining studies on the social impact of majorities and minorities, affected individuals expand their observation range and update their opinions by considering the strength, immediacy and number of both supporters and opponents. By introducing social influence and its feedback mechanism, the proposed model can highlight the heterogeneity of individuals and reproduce realistic online opinion interactions. A tradeoff is discovered between high interaction intensity and low stability with regard to observation ranges. And the distribution of individuals’ social influences presents exponential properties. The research builds a bridge to connect micro-interaction and macro-evolution of network consensus, and quantifies persuasion affects beyond local interactions. It also provides a solid theoretical and technical support for tracking opinion dynamics and analyzing network consensus in social network.4. Information competition and collision are results of the free opinion expression of users and their extensive participation in information dissemination in social network. Is not yet. At present, there is little research on the interaction and communication mode of multi competition information between home and abroad. The interaction, competition, and dissemination of multiple information yet lack of further researches. Competitive information spreading model is proposed to study interactions among multiple competitive information and their spreading process. Referring to epidemic dynamics model, the whole population is classified into four groups, and then SHIR dual competitive information diffusion model is built. We derive mean-field states transition equations to reveal the time evolution of agents with different states. Studies show that the present of hesitate status extends the interaction process, and results in an increase of dual information overall coverage. Information competition leads to a competition of neutral state agents. Advantage information dominates the whole process of information diffusion and evolution. Moreover, the ratio of final information coverage of dual information is suggested to be power-law relationship with the ratio of stable status transition probability. The research makes a connection between interactions of competitive information and diffusion of multiple information, and provides a practical and feasible research method for accurately analyzing complex multi-dimensional network consensus.
|
PDF Full Text Request